Installation

Prerequisites

  • Python 3.7 or higher
  • pip package manager

Installing ShapleyX

Standard installation (from PyPI)

pip install shapleyx

This installs the core package with all required dependencies (numpy, scipy, pandas, matplotlib, scikit-learn).

pip install shapleyx[streaming]

Adds Numba for JIT-compiled computation throughout the package. The first run compiles Numba kernels (~30--60s); subsequent runs use cached compiled code on disk. Without Numba everything still works via pure-NumPy fallbacks. Numba accelerates three independent subsystems:

Subsystem Modules Typical speedup
Streaming OMP utilities/streaming.py 4--10× on correlation scans, parallelised via prange
Surrogate prediction utilities/predictor.py 5--6× on single-sample MC Shapley evaluations
MC Shapley bootstrap utilities/mc_shapley.py 3--10× on exhaustive \(B\)-iteration bootstrap loops

With development tools

pip install shapleyx[dev]

Adds pytest, mypy, flake8, and Jupyter for local development.

From GitHub (development version)

pip install https://github.com/frbennett/shapleyx/archive/main.zip

To upgrade an existing installation:

pip install --upgrade shapleyx

Or clone and install in development mode:

git clone https://github.com/frbennett/shapleyx.git
cd shapleyx
pip install -e .

Dependencies

ShapleyX requires the following Python packages (installed automatically):

Required Optional
numpy numba — accelerates streaming OMP, surrogate prediction,
and MC Shapley bootstrap (pip install shapleyx[streaming])
scipy tqdm (progress bars during MC sampling)
pandas
matplotlib
scikit-learn

Example with the streaming extra: 

pip install shapleyx[streaming]

Verifying Installation

from importlib.metadata import version
print(f"ShapleyX v{version('shapleyx')}")